Publications

Mayrhuber E., Krauss O., “User Profile-Based Recommendation Engine Mitigating the Cold-Start Problem” in Proceedings of International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME), 2022.

Abstract

Recommendation systems can be used in many situations in daily life. Recommending people on social media networks, products in various online shops, music, or movies are only a few use cases of these systems. The cold start problem, when no information about a new or infrequent user is available, is challenging for recommendation systems. We deal with creating restaurant and category recommendations for restaurant visitors. Recommendations are generated with different metrics and technologies based on user profiles to make recommendations as individual as possible. We use k-Means and Mean-Shift for clustering users to build a base for recommendations generated using user-based and contentbased collaborative filtering methods. These suggestions consider the location of restaurants, the similarity between users and restaurants, and the ratings users give. We mitigate the cold-start problem by using matrix factorization and spatial information for users with few restaurant visits in the past. Recommendations are evaluated and adapted as a result of other user behavior to obtain better results. As a result, we can query recommendations via an Application Programming Interface (API), which consist of a mixture of location and user-based recommendation to please the users’ needs by combining exploration and exploitation.

Praschl C., Pritz S., Krauss O., Harrer M., “A Comparison Of Relational, NoSQL and NewSQL Database Management Systems For The Persistence Of Time Series Data” in Proceedings of International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME), 2022.

Abstract

Time series data is created in a variety of application areas such as sensors in cars, smartwatches or IoT devices. This kind of data is often characterized by high resource demand due to the frequency the information is measured, with data points once a day, hour and even down to milliseconds. While real-time processing of such data is often sufficient, there are also many use cases, where batch processing and consequently the storage and managed access of measurements is required. For this reason, this work evaluates different database management systems in the context of storing time related data using different data models such as classical relational models, non-relational models using NoSQL database systems and the recently upcoming group of NewSQL databases. The evaluation shows that a highly optimized time series databases such as InfluxDB is able to outperform the other tested systems regarding write-throughput and RAM as well as disk utilization in a single server setup.

Clara Diesenreiter, Oliver Krauss, Simone Sandler, Andreas Stöckl, “ProperBERT – Proactive Recognition of Offensive Phrasing for Effective Regulation” in Proceedings of International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME), 2022.

Abstract

This work discusses and contains content that may be offensive or unsettling. Hateful communication has always been part of human interaction, even before the advent of social media. Nowadays, offensive content is spreading faster and wider through digital communication channels. To help improve regulation of hate speech, we introduce ProperBERT, a fine-tuned BERT model for hate speech and offensive language detection specific to English. To ensure the portability of our model, five data sets from literature were combined to train ProperBERT. The pooled dataset contains racist, homophobic, misogynistic and generally offensive statements. Due to the variety of statements, which differ mainly in the target the hate is aimed at and the obviousness of the hate, a sufficiently robust model was trained. ProperBERT shows stability on data sets that have not been used for training, while remaining efficiently usable due to its compact size. By performing portability tests on data sets not used for fine-tuning, it is shown that fine-tuning on large scale and varied data leads to increased model portability.

Sandler Simone, Krauss Oliver, Diesenreiter Clara, Stöckl
Andreas, “Detecting Fake News and Performing Quality Ranking of German News Papers Using Machine Learning” in Proceedings of International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME), 2022.

Abstract

Nowadays, news spread quickly, and it is not always clear to the reader whether an article is real or fake. Moreover, readers use only a few sources to read the news without knowing the quality of the source. This is due to a lack of up-to-date
news or media rankings. Machine learning models can be used to automatically detect fake news. In this work, a Passive-Aggressive-Classifier, a Random-Forest, and an LSTM network are trained to distinguish between fake and non-fake (real)
news. Moreover, these models are used to classify news sources according to the amount of possible Fake News they may spread. The models are tested on English and translated German articles. The best results for Fake News detection on English articles is reached with the Passive-Aggressive-Classifier. For automatic
news ranking of translated German articles, Random-Forest provides the best result. The correlation of Random-Forest with an actual news ranking reached 0.68. This shows that automated classification can be extended to languages other than English, using this approach. In the future, other machine learning models
and translators will be used to extend the approach.

A. Veichtlbauer, C. Praschl, L. Gaisberger, G. Steinmaurer and T. I. Strasser, “Toward an Effective Community Energy Management by Using a Cluster Storage,” in IEEE Access, vol. 10, pp. 112286-112306, 2022.

DOI: 10.1109/ACCESS.2022.3216298.

Abstract

The integration of renewable local energy generation in single households – turning the household into a “prosumer” – is an important way to support an ecological transition of the electric power system. However, due to the volatile and distributed nature of most renewable energy sources, the power system may face stability problems when integrating a large number of renewables. The paper at hand describes an approach to overcome these shortages in a two-fold manner: First, the effects of the installed renewables shall be limited locally to a group of households – a so-called “energy community”. To do so, all the participating households are using existing self-consumption optimization tools. However, when a household has excess energy which can not be consumed locally, this energy is shared among the other participating households by using a cluster storage device, thus enabling a community self-consumption before feeding into the low-voltage distribution grid. Second, the connected operator may request flexibility from the participating households. For that, additional loads or load sheds are triggered by the requesting grid operator, depending on the current situation in the grid. The households decide autonomously about the amount of granted flexibility, receiving respective financial incentives. This work introduces an energy management concept and a prototypical control infrastructure used for the aforementioned functionalities. In a number of simulations and field tests, the proposed approach was successfully evaluated. The article provides a comprehensive overview of the gained results and the conclusions derived from them.

Pritz S., Praschl C., Kaiser R., Zwettler, G. “Visual Change Detection in Multi-Temporal Transects of Alpine Plants”. Proceedings of the 10th International Workshop on Simulation for Energy, Sustainable Development & Environment SESDE2022, Rom, Italy (2022).

Abstract

Due to the apparent effects of climate change on the Earth’s ecosystems, it is more important than ever to monitor flora and fauna in affected regions, e.g. mountain areas above the tree line. In the alpine ecosystem, and not just there, Vegetation plays a fundamental role and is the subject of this study. The work aims to develop algorithms for recognising small stature alpine plants from close range top view images. Ideally, automated assessment algorithms of the plant cover should objectively help scientists observe and interpret the state of the plant ecosystem over a long time series. Therefore, the aim in this respect was to derive visualisations that accurately describe plant growth and displacement (translocation). Additionally, recording changes in biodiversity was an intent. This work uses multi-temporal data comprising RGB images and multi-label masks to accomplish the aforementioned task. The evaluated methods involve mask comparison, optical flow estimation, detection of individual plants, and descriptive statistical analysis of image feature properties. Tests on the given data set show that all methods but the optical flow estimation have great potential. The mask comparison method captured plant growth and translocation most satisfactory. Individual plant detection and statistical analysis further helped to evaluate changes in biodiversity. When combined, the proposed methods give an immediate overview about relevant changes in the multi-temporal transects, which has not been done before for close-distance images of alpine plants.

Krauss O., Aschauer A., Stöckl A. “Modelling shifting trends over time via topic analysis of text documents”. Proceedings of the 34rd European Modeling and Simulation Symposium EMSS2022, Rom, Italy (2022).

Pointner A., Praschl C., Krauss O. “Towards Modelling Namespaces in Graph Databases”. Proceedings of the 34rd European Modeling and Simulation Symposium EMSS2022, Rom, Italy (2022).

Abstract

We present a novel approach to store data with different contexts inside a property graph model. We introduce namespaces, similar to namespaces in XML, and extend nodes and relationships with labels to assign them to a specific context, i.e. namespace. Individual properties of a node or relationship can also be put in a namespace. This work is specifically targeting the utilization in graph databases, with a reference implementation provided via the Neo4j database. In addition to the theoretical approach, an object to graph mapper for the programming language Java is implemented and used to evaluate the approach. As an evaluation example, a university organization is used, which is split into two domains. The experiments show, that information of different domains can be stored within the same model using namespaces. Thus, it is possible to reuse shared information over multiple contexts, which reduces data duplication in the graph database, as otherwise multiple nodes would be required.

Zwettler, G., Ono, Y., Stradner, M., & Praschl, C. “Strategies for Semi-Automated Registration of Historic Aerial Photographs Utilizing Street and Roof Segmentations as Durable Landmarks”. Proceedings of the 34th European Modeling and Simulation Symposium EMSS2022, Rom, Italy (2022).

Abstract

Historical and current aerial photographs are only of great value if the geolocation or address of the photographed areas is also available. In Western Europe, especially Austria, Germany and Czech Republic, there is a market for the sale of aerial photographs of one’s own private residential building. Automated geolocation is a feasible way to enable the sales agents to assign the addresses for the sale more quickly. In the course of this research work, a process chain is modeled that allows the assignment of aerial photographs to residential addresses using machine vision. After model-based rectifying the aerial images to compensate for perspective distortions, larger image blocks get assembled using image stitching. The assignment to a 2D reference map, such as satellite imagery via Google Maps, is done by applying a U-Net CNN after extracting durable image features such as roads or buildings. The mapping of aerial imagery to two-dimensional cartography is either automated via registration approaches or based on manually placed corresponding landmarks and homography. Test runs on imagery between the years 1969 and 2020 show that the labor-intensive process of geolocation of aerial imagery can be solved by the proposed process model in a hybrid way.

Meindl R., Sandler S., Mayrhuber E., and Krauss O. “Distributed Classification – A Scalable Approach to Semi SupervisedMachine Learning” Proceedings of the 34th European Modeling and Simulation Symposium EMSS2022, Rom, Italy (2022).

Abstract

Fitting real world data into a model for classification, is a challenging task. Modern approaches to classification are often resource intensive and may become bottlenecks. A microservice architecture that allows maintaining a model of real world data, and adding new information as it becomes available is presented in this paper. Updates to the model are handled via different microservices. The architecture and connected workflows are demonstrated in a use case of classifying text data in a taxonomy represented by a directed acyclic graph (DAG). The presented architecture removes the classification bottleneck, as multiple data points can be added independent of each other, and reading access to the model is not restricted. Additional microservices also enable a manual intervention to update the model.

Praschl C., Pointner A., Krauss O., Helm E., Schuler A. “Model Verification in Graph Databases and its Application in Neo4j.” Proceedings of the 34th European Modeling and Simulation Symposium EMSS2022, Rom, Italy (2022).

Abstract

This work introduces a concept for rule based model verification using a graph database on the example of Neo4j and its query language Cypher. An approach is provided that allows to define verification rules using a graph query language to detect transformation errors within a given domain model. The approach is presented based on a running example, showing its capability of detecting randomly generated errors in a transformation process. Additionally, the method’s performance is evaluated using multiple subsets of the IMDb movie data with a maximum of 17,000,000 nodes and 41,000,000 relationships. This performance evaluation is carried out in comparison to the Object Constraint Language, showing advantages in the context of highly connected datasets with a high number of nodes. Another benefit is the utilization of a well established graph database as verification tool without any need for re-implementing graph and pattern matching logic.

Kaiser R., Praschl C., Zwettler G. “Long-Term Monitoring of Alpine Plant Diversity in the National Park Hohe Tauern”. 7. Symposium for research in protected areas. Conference. Vol. 2. 2022.

Abstract

The Hohe Tauern National Park has founded an interdisciplinary monitoring and research program for long-term observation of alpine ecosystems. This initiative provides – among other findings – an ongoing digital image archive in the form of strictly standardized (geo-static, colourfast), high-resolution (1 px. ≈ 0,1mm) nadir photos (view vertical to the ground) with high information content and great relevance in terms of documentation. These data, comparable to earth orthophotos, represent the basis for the project at hand. It focuses on developing a software prototype to automatically recognize plants from image data using computer vision and machine learning. The goals are threefold. First, the reliable recognition of individual plant species and their individuals, despite overlap with other plants or vegetation structures, is aimed. Secondly, the variation in nature and thus divergent appearance of a specimen is addressed. Thirdly, it should be possible to detect identical plants within a time series. In addition, this should allow the models to be updated when new data are available.

Pointner, A., Spitzer, EM., Krauss, O., Stöckl, A. (2023). Anomaly-Based Risk Detection Using Digital News Articles. In: Arai, K. (eds) Intelligent Systems and Applications. IntelliSys 2022. Lecture Notes in Networks and Systems, vol 542. Springer, Cham.

DOI: https://doi.org/10.1007/978-3-031-16072-1_1

Abstract

Enterprise risk management is a well established methodology used in industry. This area relies heavily on risk owners and their expert opinion. In this work, we present an approach to a semi-automated risk detection for companies using anomaly detection. We present various anomaly detection algorithms and present an approach on how to apply them on multidimensional data sources like news articles and stock data to automatically extract possible risks. To do so, NLP methods, including sentiment analysis, are used to extract numeric values from news articles, which are needed for anomaly analysis. The approach is evaluated by conducting interview questionnaires with domain experts. The results show that the presented approach is a useful tooling that helps risk owners and domain expert to find and detect potential risks for their companies.

Fernandez-Llatas, Carlos, et al. “Building Process-Oriented Data Science Solutions for Real-World Healthcare.” International Journal of Environmental Research and Public Health 19.14 (2022): 8427.

DOI: https://doi.org/10.3390/ijerph19148427

Abstract

The COVID-19 pandemic has highlighted some of the opportunities, problems and barriers facing the application of Artificial Intelligence to the medical domain. It is becoming increasingly important to determine how Artificial Intelligence will help healthcare providers understand and improve the daily practice of medicine. As a part of the Artificial Intelligence research field, the Process-Oriented Data Science community has been active in the analysis of this situation and in identifying current challenges and available solutions. We have identified a need to integrate the best efforts made by the community to ensure that promised improvements to care processes can be achieved in real healthcare. In this paper, we argue that it is necessary to provide appropriate tools to support medical experts and that frequent, interactive communication between medical experts and data miners is needed to co-create solutions. Process-Oriented Data Science, and specifically concrete techniques such as Process Mining, can offer an easy to manage set of tools for developing understandable and explainable Artificial Intelligence solutions. Process Mining offers tools, methods and a data driven approach that can involve medical experts in the process of co-discovering real world evidence in an interactive way. It is time for Process-Oriented Data scientists to collaborate more closely with healthcare professionals to provide and build useful, understandable solutions that answer practical questions in daily practice. With a shared vision, we should be better prepared to meet the complex challenges that will shape the future of healthcare.

Praschl C., Thiele E., Krauss O. “Utilization of Geographic Data for the Creation of Occlusion Models in the Context of Mixed Reality Applications”. Extended Reality, 1st ed. Lecture Notes in Computer Science. Volume 13446. (2022).

DOI: https://doi.org/10.1007/978-3-031-15553-6

Abstract

Emergency responder training can benefit from outdoor use of Mixed Reality (MR) devices to make trainings more realistic and allow simulations that would otherwise not be possible due to safety risks or cost-effectiveness. But outdoor use of MR requires knowledge of the topography and objects in the area to enable accurate interaction of the real world trainees experience and the virtual elements that are placed in them. An approach utilizing elevation data and geographic information systems to create effective occlusion models is shown, that can be used in such outdoor training simulations. The initial results show that this approach enables accurate occlusion and placement of virtual objects within an urban environment. This improves immersion and spatial perception for trainees. In the future, improvements of the approach are planned with on the fly updates to outdated information in the occlusion models.

Callan, James, et al. “How do Android developers improve non-functional properties of software?.” Empirical Software Engineering 27.5 (2022): 1-42.

DOI: https://doi.org/10.1007/s10664-022-10137-2

Abstract

Nowadays there is an increased pressure on mobile app developers to take non-functional properties into account. An app that is too slow or uses much bandwidth will decrease user satisfaction, and thus can lead to users simply abandoning the app. Although automated software improvement techniques exist for traditional software, these are not as prevalent in the mobile domain. Moreover, it is yet unknown if the same software changes would be as effective. With that in mind, we mined overall 100 Android repositories to find out how developers improve execution time, memory consumption, bandwidth usage and frame rate of mobile apps. We categorised non-functional property (NFP) improving commits related to performance to see how existing automated software improvement techniques can be improved. Our results show that although NFP improving commits related to performance are rare, such improvements appear throughout the development lifecycle. We found altogether 560 NFP commits out of a total of 74,408 commits analysed. Memory consumption is sacrificed most often when improving execution time or bandwidth usage, although similar types of changes can improve multiple non-functional properties at once. Code deletion is the most frequently utilised strategy except for frame rate, where increase in concurrency is the dominant strategy. We find that automated software improvement techniques for mobile domain can benefit from addition of SQL query improvement, caching and asset manipulation. Moreover, we provide a classifier which can drastically reduce manual effort to analyse NFP improving commits.

Meindl, Rainer, et al. “A Scalable Microservice Infrastructure for Fleet Data Management.” International Conference on Database and Expert Systems Applications. Springer, Cham, 2022.

DOI: https://doi.org/10.1007/978-3-031-14343-4_37

Abstract

Modern Internet of Things solutions using edge devices produce large amounts of raw data. In order to utilize this data, it needs to be processed, aggregated, and categorized to enable decision making for management and end-users. This data management is a non-trivial task, as the computational load is directly proportional to the amount of data. In order to tackle this issue, we provide an extensible and scalable microservice architecture that can receive, normalize, and filter the raw data and persist it in different levels of aggregation, as well as for time series analysis.

Spitzer, Eva-Maria, Oliver Krauss, and Andreas Stöckl. “Accurately Predicting User Registration in Highly Unbalanced Real-World Datasets from Online News Portals.” International Conference on Database and Expert Systems Applications. Springer, Cham, 2022.

DOI: https://doi.org/10.1007/978-3-031-12423-5_23

Abstract

Getting visitors to register is a crucial factor in marketing for online news portals. Current approaches are rule-based by awarding points for specific actions [3]. Finding efficient rules can be challenging and depends on the specific task. Registration is generally rare compared to regular visitors, leading to highly imbalanced data.

We analyze different supervised learning classification algorithms under consideration of the data imbalance. As case study, we use anonymized real-world data from an Austrian newspaper outlet containing the visitor’s session behavior with around 0.1% registrations over all visits.

We identify an ensemble approach combining the Balanced Random Forest Classifier and the RUSBoost Classifier correctly identifying 76% of registrations over five independent data sets

Praschl C., Stradner M., Ono Y., Zwettler G. “Towards an Automated System for Reverse Geocoding of Aerial Photographs”. WSCG 2022: proceedings: 30. International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision, p. 296-301. (2022).

DOI: https://www.doi.org/10.24132/CSRN.3201.37

Abstract

Aerial photographs of buildings are often used as memorabilia sold by trading companies. Such photographs come with an issue regarding the address of the shown buildings, since the recording location of the camera may be known, but shows a spatial distance to the actual subject of the image. In addition to that, also this recording location is often not known in detail but only roughly in the form of the flight route/area. To address this problem, a methodology for reverse geocoding is proposed, allowing to identify the position of buildings that are photographed from aerial vehicles. This is done using a process for extending recording locations and a second process based on the registration of invariant features within aerial shots compared to maps.

Praschl C., Krauss O. “Geo-Referenced Occlusion Models for Mixed Reality Applications using the Microsoft HoloLens.” VISIGRAPP (3: IVAPP). 2022.

DOI: http://dx.doi.org/10.5220/0010775200003124

Abstract

Emergency responders or task forces can benefit from outdoor Mixed Reality (MR) trainings, as they allow more realistic and affordable simulations of real-world emergencies. Utilizing MR devices for outdoor situations requires knowledge of real-world objects in the training area, enabling the realistic immersion of both, the real, as well as the virtual world, based on visual occlusions. Due to spatial limitations of state-of-the-art MR devices recognizing distant real-world items, we present an approach for sharing geo-referenced 3D geometries across multiple devices utilizing the CityJSON format for occlusion purposes in the context of geospatial MR visualization. Our results show that the presented methodology allows accurate conversion of occlusion models to geo-referenced representations based on a quantitative evaluation with an average error according to the vertices’ position from 1.30E-06 to 2.79E-04 (sub-millimeter error) using a normalized sum of squared errors metric. In the future, we plan to also incorporate 3D reconstructions from smartphones and drones to increase the number of supported devices for creating geo-referenced occlusion models. 

Egelkraut, Reinhard, et al. “Open Infrastructure for Standardization of HL7® FHIR® Implementation Guides in Austria.” dHealth. 2022.

DOI: https://doi.org/10.3233/shti220372

Abstract

Background: HL7 Austria is a non-profit association dedicated to improving electronic data communication and interoperability in healthcare using the HL7 international standards. Objectives: We aim to provide an open infrastructure to develop, manage, and maintain HL7 FHIR implementation guides. Methods: We utilize state-of-the-art open-source tooling developed by the FHIR community to support continuous integration. Results: The implementation guides can be published as static HTML websites and maintained using GitHub. Conclusion: The solution supports all steps of a standard’s lifecycle, from drafting and reviewing to balloting, publishing, and maintenance.

Praschl C. Auserperg-Castell P., Forster-Heinlein B., Zwettler G. (2021). Segmentation and Multi-Facet Classification of Individual Logs in Wooden Piles. In Computer Aided Systems Theory Extended Abstract.

Link: https://eurocast2022.fulp.ulpgc.es/sites/default/files/Eurocast_2022_Extended_Abstract_Book.pdf

Abstract

The inspection of products and assessment of quality is connected with high costs and time effort in many industrial domains. This also applies to the forestry industry. Utilizing state-of-the-art deep learning models allows automizing the analysis of wooden piles in a vision-based manner. In this work a parallel two-step approach is presented for the segmentation and multi-facet classification of individual logs, according to the wood type and quality. The present approach is based on a preliminary log localization step and like this allows determining the quality, volume and also the value of individual logs, respectively the whole wooden pile. Using a YOLOv4 model for wood species classification for douglas firs, pines and larches results in an accuracy of 74.53%, while a quality classification model for spruce logs reaches 86.58%. In addition to that, the trained U-NET segmentation model reaches an accuracy of 93%. In the future, the underlying data set and models will be further improved and integrated to a mobile application for the on site analyzation of wooden piles by foresters.

Jorge Munoz-Gama, Niels Martin, Carlos Fernandez-Llatas, Owen A. Johnson, Marcos Sepúlveda, Emmanuel Helm, Victor Galvez-Yanjari, Eric Rojas, Antonio Martinez-Millana, Davide Aloini, Ilaria Angela Amantea, Robert Andrews, Michael Arias, Iris Beerepoot, Elisabetta Benevento, Andrea Burattin, Daniel Capurro, Josep Carmona, Marco Comuzzi, Benjamin Dalmas, Rene de la Fuente, Chiara Di Francescomarino, Claudio Di Ciccio, Roberto Gatta, Chiara Ghidini, Fernanda Gonzalez-Lopez, Gema Ibanez-Sanchez, Hilda B. Klasky, Angelina Prima Kurniati, Xixi Lu, Felix Mannhardt, Ronny Mans, Mar Marcos, Renata Medeiros de Carvalho, Marco Pegoraro, Simon K. Poon, Luise Pufahl, Hajo A. Reijers, Simon Remy, Stefanie Rinderle-Ma, Lucia Sacchi, Fernando Seoane, Minseok Song, Alessandro Stefanini, Emilio Sulis, Arthur H.M. ter Hofstede, Pieter J. Toussaint, Vicente Traver, Zoe Valero-Ramon, Inge van de Weerd, Wil M.P. van der Aalst, Rob Vanwersch, Mathias Weske, Moe Thandar Wynn, Francesca Zerbato. Process Mining for Healthcare: Characteristics and Challenges, Journal of Biomedical Informatics. 2022.

DOI: https://doi.org/10.1016/j.jbi.2022.103994

Abstract

Process mining techniques can be used to analyse business processes using the data logged during their execution. These techniques are leveraged in a wide range of domains, including healthcare, where it focuses mainly on the analysis of diagnostic, treatment, and organisational processes. Despite the huge amount of data generated in hospitals by staff and machinery involved in healthcare processes, there is no evidence of a systematic uptake of process mining beyond targeted case studies in a research context. When developing and using process mining in healthcare, distinguishing characteristics of healthcare processes such as their variability and patient-centred focus require targeted attention. Against this background, the Process-Oriented Data Science in Healthcare Alliance has been established to propagate the research and application of techniques targeting the data-driven improvement of healthcare processes. This paper, an initiative of the alliance, presents the distinguishing characteristics of the healthcare domain that need to be considered to successfully use process mining, as well as open challenges that need to be addressed by the community in the future. 

C. Praschl, O. Krauss. Geo-Referenced Occlusion Models for Mixed Reality Applications Using the Microsoft HoloLens. Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, Volume 3: IVAPP. 2022.

DOI: https://doi.org/10.5220/0010775200003124

Abstract

Emergency responders or task forces can benefit from outdoor Mixed Reality (MR) trainings, as they allow more realistic and affordable simulations of real-world emergencies. Utilizing MR devices for outdoor situations requires knowledge of real-world objects in the training area, enabling the realistic immersion of both, the real, as well as the virtual world, based on visual occlusions. Due to spatial limitations of state-of-the-art MR devices recognizing distant real-world items, we present an approach for sharing geo-referenced 3D geometries across multiple devices utilizing the CityJSON format for occlusion purposes in the context of geospatial MR visualization. Our results show that the presented methodology allows accurate conversion of occlusion models to geo-referenced representations based on a quantitative evaluation with an average error according to the vertices’ position from 1.30E-06 to 2.79E-04 (sub-millimeter error) using a normalized sum of squared errors metric. In the future, we plan to also incorporate 3D reconstructions from smartphones and drones to increase the number of supported devices for creating geo-referenced occlusion models.

C. Praschl, G. Zwettler. Three-Step Approach for Localization, Instance Segmentation and Multi-Facet Classification of Individual Logs in Wooden Piles. Proceedings of the 11th International Conference on Pattern Recognition Applications and Methods. 2022.

DOI: https://doi.org/10.5220/0010892100003122

Abstract

The inspection of products and the assessment of quality is connected with high costs and time effort in many industrial domains. This also applies to the forestry industry. Utilizing state-of-the-art deep learning models allows the analysis automation of wooden piles in a vision-based manner. In this work a three-step approach is presented for the localization, segmentation and multi-facet classification of individual logs based on a client/server architecture allowing to determine the quality, volume and like this the value of a wooden pile based on a smartphone application. Using multiple YOLOv4 and U-NET models leads to a client-side log localization accuracy of 82.9% with low storage requirements of 23 MB and a server-side log detection accuracy of 94.1%, together with a log type classification accuracy of 95% and 96% according to the quality assessment of spruce logs. In addition, the trained segmentation model reaches an accuracy of 89%.

Helm, Emmanuel, et al. “FHIR2BPMN: Delivering Actionable Knowledge by Transforming Between Clinical Pathways and Executable Models.” Healthcare of the Future 2022. IOS Press, 2022. 9-14.

DOI: https://doi.org/10.3233/SHTI220311

Abstract

Healthcare processes have many particularities captured and described within standards for medical information exchange such as HL7 FHIR. BPMN is a widely used standard to create readily understandable processes models. We show an approach to integrate both these standards via an automated transformation mechanism. This will allow us to use the various tools available for BPMN to visualize and automate processes in the healthcare domain. In the future we plan to extend this approach to enable mining and analyzing executed processes.

IGSOFT Softw. Eng. Notes 44, 3 (July 2019) (Details)
William B. Langdon, Westley Weimer, Christopher Timperley, Oliver Krauss, Zhen Yu Ding, Yiwei Lyu, Nicolas Chausseau, Eric Schulte, Shin Hwei Tan, Kevin Leach, Yu Huang, and Gabin An

arXiv preprint arXiv:1907.03773

Proceedings of the VISAPP 2020, Valetta, Malta, 2020, pp. 12
G. Zwettler, D. Holmes III, W. Backfrieder

Process-Oriented Data Science for Healthcare (Details)
Emmanuel Helm, David Baumgartner, Anna M. Lin, Alvin Lin, Josef Küng

Proceedings of the 6th International Workshop on Genetic Improvement
Oliver Krauss, Hanspeter Mössenböck, Michael Affenzeller

ICT for Health Science Research
A. Lin, O. Krauss, E. Helm

Process Mining Conference 2019 – 1st International Conference on Process Mining, June 24-26, 2019, Aachen, Germany
Baumgartner D., Haghofer A., Limberger M., Helm E.

Information Systems and Neuroscience, p. 221 – 228, Springer Verlag
Baumgartner D., Fischer T. Riedl R., Dreiseitl S.